Contextual Speech Recognition for Virtual Meeting Jargon

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Virtual meeting systems often fail to accurately transcribe company-specific or team-specific jargon due to unrecognized words, leading to inaccurate text generation in closed-captions, transcripts, and summaries.

Innovation Solution

A system that modifies a speech recognition system by identifying and incorporating previously unrecognized content, such as names, documents, and shared text during meetings, using optical character recognition and machine learning models to generate accurate transcriptions and summaries.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a speech recognition system uses general training data, then it can process diverse speech patterns, but it fails to accurately transcribe company-specific or team-specific jargon

Engineering Contradiction:
Improvetranscription accuracyVSAvoidability to recognize specialized terminology
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The system performs preliminary actions by extracting text from meeting content items (documents, slides, chat) before the speech recognition process. This extracted text is then used to pre-train or fine-tune the speech recognition model specifically for the meeting context, enabling it to accurately transcribe specialized jargon and terminology that appears in the meeting materials.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system uses feedback from the meeting content items to continuously improve the speech recognition accuracy. By extracting text from documents, slides, and chat messages, and using this feedback to train or adjust the speech recognition model, the system iteratively improves its ability to transcribe accurate terminology specific to that meeting context.

Inventive Principle:
Principle #23Feedback

2Measurement precision

If the speech recognition system is modified to recognize previously unrecognized content, then transcription accuracy improves, but system complexity increases

Engineering Contradiction:
Improveword recognition precisionVSAvoidsystem modification complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the speech recognition task into distinct components: (1) extracting text from content items, (2) identifying previously unrecognized words, (3) training or fine-tuning the speech recognition model with this new data, and (4) generating the final transcription. This segmentation allows each component to be optimized independently and simplifies the overall system architecture.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system introduces an intermediary process that extracts text from meeting content items and uses this as bridging data to train the speech recognition model. This intermediary text extraction and pre-training step acts as a mediator between the raw audio input and the final transcription output, improving precision without requiring direct complex modifications to the core speech recognition algorithm.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Reliability

If the system processes all content items to modify the speech recognition system, then transcription accuracy improves, but processing time increases

Engineering Contradiction:
Improvetranscription reliabilityVSAvoidprocessing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system applies partial action by selectively processing only the most relevant content items for training the speech recognition model. Instead of processing every single content item exhaustively, the system identifies and processes a representative sample or the most impactful content items that contain the most valuable terminology and context for improving transcription accuracy.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4693278A1Contextual speech recognition of virtual meetings
Publication Date: 2026.02.11 GOOGLE LLC
  • EP4693278A1 patent drawingFigure 1
  • EP4693278A1 patent drawingFigure 2
  • EP4693278A1 patent drawingFigure 3

AI summary

A method includes receiving audio data of a virtual meeting and identifying, within a plurality of content items related to the virtual meeting, content not previously recognized by a speech recognition system designated to convert the audio data of the virtual meeting into text. The method also includes causing the speech recognition system to be modified based on the previously unrecognized content. The method further includes causing the audio data of the virtual meeting to be converted into the text using the modified speech recognition system, wherein the text comprises at least part of the previously unrecognized content.